Gaussian noise removal in gray scale images using fast Multiscale Directional Filter Banks

Abstract

This paper presents a novel approach for Gaussian noise removal using Multiscale Filter Banks for the Contourlet Transform. The Multiscale Directional Filter Bank (MDFB) improves the radial frequency resolution of the Contourlet Transform by introducing an additional decomposition in the high frequency band. This reduces the computational complexity significantly by saving a directional decomposition because of the change in the order of decomposition. Scaling is performed by a low pass filtering based splitting and the scale decomposition is done by the Directional Filter Bank. Perfect reconstruction is possible for the scale decomposition regardless of the choice of the low pass filter. MDFB outperforms the conventional Wavelet and Contourlet transform methods for Gaussian noise removal. Denoising performance of this proposed method is compared with Wavelet and Contourlet based denoising schemes with state of art threshold methods.